Skip to Content
Your Clinical Trial Deserves TLC From PRC

Predicting the Success of Clinical Trials

April 5, 2017

By Angela Woodall, Healthcare and Environmental Reporter.

Every so often a drug study catches the attention of patients, investors, and the business press. The late phase trial involving Stimuvax, a vaccine therapy for lung cancer, definitely counts as one of those. Specialists in the medical community were uncharacteristically hopeful, an oncologist wrote several years later. But financial analysts and others zeroed in on the trial, many of them all but convinced that the success of the phase II Stimulax trial was “a foregone conclusion” as a team of researchers put it.

The team was looking at strategies for predicting the success of clinical trials, which can be hindered by commercial viability, safety, or efficacy depending on trial stage.  This study was one of a growing number of analyses that use data to find what separates therapies that reach approval from the ones that don’t reach the finish line, even borrowing from baseball’s “Moneyball” sabermetrics to do it.

In science, a negative trial result isn’t always a bad thing. However negative results can be expensive and time-consuming. Predicting the probability that a trial will be successful provides a strong incentive to pull together stakeholders who don’t always stand under the same umbrella. They include trial sponsors, their competitors, investigators, and patients, in addition to venture capitalists and stockbrokers.

The stakeholders would like to improve the success rates for drug development which are stubbornly low. Oncology drugs in phase III trials, for example, have less than a 50%chance of approval according to researchers at Tufts Center for the Study of Drug Development and Janssen Research and Development. From the glass-half-full perspective, the rate is a sign that the system is working well at keeping potentially unsafe therapies off the market. However, stakeholders have been shooting in the dark when it comes to selecting which candidates to back according to the Tufts-Janssen teamwho proposed to improve the odds with an algorithm that assesses the probability of success for regulatory marketing approval after completion of phase II testing. The number of participants enrolled in phase II was one of the four most significant predictors they found: “Having a relatively large number of patients enrolled in pivotal phase II clinical trials reduces uncertainty about the results.”

A third study used sabermetrics, a school of baseball statistics that captured the imagination of Hollywood in the film “Instead of Major League Baseball, three researchers used the approach to analyze overlooked features in trials to predict the likelihood a trial would fail because of toxicity reasons.

Other attempts mined the scientific literature. By searching for patterns in publications detailing clinical trial findings, one group found that publication count and the dedication of experts over time involved in the trials made a difference. A high failure rate in drug development, in oncology at least, highlights the severe limitations of prediction methods according to a team who find commonalities associated with FDA approval.

Study design can also play a part in success. Adaptive designs, according to a 2014 paper, allow more efficient use of information for decision making. This ultimately translates into improved probability of success and shorter overall time to market for successful products. They bring with them some distinctive challenges. Alternatively, the FDA has taken a friendly stance toward adaptive trial designs as a way to potentially help open up the biopharmaceutical pipeline.

Yet others are applying machine learning techniques to improve participation in trials, a challenge that can “compromise results or stop some studies altogether.” Computational methods are also being applied to site performance. Under tighter budgets and timelines, pharmaceutical firms, biotechnology companies, and CROs are pouring over reams of descriptive and historical performance data to identify factors that better predict successful patient enrollment,. The four factors that stood out involved past performance, experience, investigative site focus, and historic speed to randomize the first study volunteer. For example, CROs and sponsors that have spent time analyzing site performance datasets have found that once a particular site has conducted 6 to 10 clinical trials, that site has a higher likelihood of meeting enrollment targets within the requisite time frame, Getz wrote.

All the examples come with the caveat that the universe of possibilities they are basing predictions on is a closed one. In other words, uncertainty is a constant companion but methods do give sponsors potential tools for deciding which novel therapies (or devices) to back.

Angela Woodall, Healthcare and Environmental Reporter

Back to Blog

Receive Our Next News and Blog Posts

Subscribe to our monthly news!

[mc4wp_form id="6345"]

Does Your Trial Need TLC From PRC?

Let’s Talk